NASA LEARN Program · • Conference paper presented at the 2013 AIAA GNC Conf. • Conference paper submitted to the 2014 ACC (details in next page). ... (Virtual Leader). • Achieved

Grant Number: NNX13AB74A

Flight Control Systems Lab (FCSL), Interactive Robotics Lab (IRL)

Department of Mechanical and Aerospace Engineering,

West Virginia University (WVU)

NASA LEARN Program

November 14, 2013, NASA LEARN Seminar

Cooperative Gust Sensing and Suppression

for Aircraft Formation Flight

Nov. 16, 2012

Aviation Safety Research and Development

November 14, 2013

Cooperative Gust Sensing & Suppression

Outline

I. Project Overview & Status

II. Formation Flight Simulator

III. Experimental Flight Validation

IV. Gust/Wake Sensing and Suppression Control

V. Conclusions & Plan for Future Research

2

Nov. 16, 2012


November 14, 2013


Objectives & Challenges

3

NASA LEARN Project Phase I:

“Cooperative Gust Sensing & Suppression for Aircraft

Formation Flight”

PI: Dr. Marcello Napolitano, Co-PI: Dr. Yu Gu (West Virginia Univ.)

Co-PI: Dr. Haiyang Chao (currently at University of Kansas)

Objectives Development of a cooperative strategy for gust sensing and

suppression within a formation flight.

Tasks 1. Cooperative Gust and Turbulence Sensing and Prediction;

2. Flight Simulation and Validation;

3. Active Gust Suppression Control.

Nov. 16, 2012


November 14, 2013


Project Introduction

4

Innovation: 1. Use of a formation flying sub-scale aircraft for wind/wake

sensing and estimation;

2. A cooperative gust/wake estimation and suppression control

strategy that uses the leader aircraft as a remote wind sensor.

Impact: 1. Enabling the technology for routine formation flight (toward

environmentally responsible aviation operations);

2. Similar gust sensing and suppression control algorithms could

be introduced on commercial aircraft with the goals of reducing

gust load and increasing passenger comfort;

3. Providing experimental validation of wake models.

Nov. 16, 2012


November 14, 2013


Project Status

5

Videos of the WVU close formation flight experiments available at:

https://www.youtube.com/watch?v=3tKVDRj0UYw

https://www.youtube.com/watch?v=LssOqx9knIY





Nov. 16, 2012


November 14, 2013


Deliverables to date

6

1. Gust estimation method from onboard sensor measurements;

• Conference paper presented at the 2013 AIAA GNC Conf.

• Conference paper submitted to the 2014 ACC (details in next page).

2. Preliminary set of active gust suppression control laws;

• Preparation for a conference submission to AIAA GNC 2015.

3. Comprehensive formation flight simulator;

• Conference paper submitted to ACC 2014.

4. Multi-UAV framework including both hardware and software to

support further application of formation flight test;

• Conference paper submitted to ACC 2014.

5. Project website: http://www2.statler.wvu.edu/~irl/page16.html

http://www2.statler.wvu.edu/~irl/page16.html

http://www2.statler.wvu.edu/~irl/page16.html

Nov. 16, 2012


November 14, 2013


Publications

7

1. Matthew Rhudy, Mario L. Fravolini, Yu Gu, Marcello R. Napolitano, Srikanth Gururajan, and

Haiyang Chao, “Cross-Platform Evaluation of UKF Airspeed Estimation using UAV Flight

Data”, submitted to IEEE Transactions on Aerospace and Electronic Systems, currently under

review.

2. Matthew Rhudy, Trenton Larrabee, Haiyang Chao, Yu Gu, and Marcello R. Napolitano, “UAV

Attitude, Heading, and Wind Estimation Using GPS/INS and an Air Data System”, AIAA

Guidance, Navigation, and Control Conference, August, 2013.

3. Trenton Larrabee, Haiyang Chao, Yu Gu, and Marcello R. Napolitano, “Wind field and wake

estimation in UAV formation flight”, submitted to the 2014 American Control Conference,

currently under review.

4. Caleb Rice, Yu Gu, Haiyang Chao, Trenton Larrabee, Srik Gururajan, and Marcello R.

Napolitano, “Control performance analysis for autonomous close formation flight

experiments”, submitted to the 2014 American Control Conference, currently under review.

Nov. 16, 2012


November 14, 2013


Project Personnel

8

Faculty Members

PI: Dr. Marcello Napolitano, WVU Professor.

Co-PI: Dr. Haiyang Chao, WVU Post-Doc (Jan. – July 2013), Assistant

Professor at KU (Aug.2013 - present).

Co-PI: Dr. Yu Gu, WVU Assistant Professor

Students

- Trenton Larabee, M.S. Student (Task: gust and wake sensing

algorithm design, simulation, and flight testing);

- Caleb Rice, M.S. Student (Task: formation flight controller

implementation, simulation, and flight testing);

- Lucas Behrens, UG Senior (Task: UAV maintenance, and support of

flight testing activities).

Nov. 16, 2012


November 14, 2013


Technical Approach

9

Formation Flight

1. Design of outer loop flight controller (GPS Trajectory tracking);

2. Design of formation flight controller;

Gust & Wake Sensing

1. Gust sensing algorithm design;

2. Cooperative Gust/Wake sensing algorithm design;

Gust Alleviation Control

1. Design of gust suppression controller;

2. Simulation using “Phastball” mathematical model.

Flight Test Validation

1. Formation flight of two WVU “Phastball” aircraft;

2. Gust/wake sensing with different offsets for formation flight.

Nov. 16, 2012


November 14, 2013


Outline




IV. Gust/Wake Sensing and Suppression

V. Conclusions & Plans for Future Research

10

Nov. 16, 2012


November 14, 2013


WVU “Phastball” Sub-Scale Research Aircraft

11

Specifications

Length ~ 88 in. (~2.2m)

Wing Span ~ 96 in. (~2.4m)

Take-off Weight ~ 26 lbs. (~12 Kg)

Max Payload ~ 7 lbs. (~3.2 Kg)

Thrust

(Ducted Fan)

2 × 25 N

Flight Duration ~ 7 min.

Cruise Speed ~ 30 m/s

Controls Elev., Ail., Rud.,

Throttle

Nov. 16, 2012


November 14, 2013


WVU Gen-V Avionics

12

Key Features

•9 Controllable Channels

•GPS/IMU/MAG

•8ch. PWM Pilot Input

•12ch., 16-bit Aux A/D

•“Black-Box” Recorder

•RF Modem Up/Downlink

•400m Laser Range Finder

•Alpha/Beta Vanes

•Pitot-Tube for Static/Dynamic

Pressure

•Humidity and Temperature

Nov. 16, 2012


November 14, 2013


“Phastball” Simulator (cont.)

• WVU Simulink-based “Phastball” simulator (using FDC package)

(FDC model is modified based on PID-derived aerodynamic coefficients)

Nov. 16, 2012


November 14, 2013


Formation Flight Simulator

• WVU “Phastball” Formation Flight Simulator with 3-aircraft Formation

– Outer loop:

» Vertical controller: based on PD type controller.

» Horizontal controller: based on Non Linear Dynamic Inversion (NLDI)

– Inner loop (attitude tracking): based on longitudinal/lateral LQ design

Autonomous Formation Flight – Design and Experiments, Y. Gu, G. Campa, M. Napolitano, et. al., InTech, Chapter 2009.

Nov. 16, 2012


November 14, 2013


Formation Flight Controller

15

Overview of NLDI based Formation Flight Controller

UAV

Dynamics

C Throttle

Elevator

Aileron

LLH

Attitude

Gyro/Flow Angles

LQR

(lon)

LQR

(lat)

Trim

Trim

+

+

+

+

Roll

Pitch

Pitch_ref

Roll_ref

Rudder

Trim+

+

-

+

+

-

q/α

p/r/β

Nov. 16, 2012


November 14, 2013


Wake Modeling and Simulation

• Hallock-Burnham vortex: 𝑣𝜃 𝑟 =Γ𝑖

2𝜋𝑟

𝑟2

𝑟2+𝑟𝑐2

• Sarpkaya wake delay model: Γ𝑖 = Γ0exp−𝐶𝑑(𝜀Γ0)

0.25

1.2727𝑉0𝑏0

Wake vortices of Phastball UAV after roll-up (Core radius: 0.09 m, initial circulation: 1.72 m2/s)

Nov. 16, 2012


November 14, 2013


Formation Flight Using “Phastball” Model (cont.)

• Three Aircraft Formation Simulation

– Using RC flight data as the leader aircraft (Red).

– Simulating followers (Blue/Green – 20/20/20 m).

-500

-400

-300

-200

-100

0

100

200

300

400

500

-300

-200

-100

0

100

200

300

-300

-200

-100

0

100

200

300

z u

p (

m)

t: 140.000000 s

x north (m)y east (m)

Nov. 16, 2012


November 14, 2013


Outline






18

Nov. 16, 2012


November 14, 2013


WVU Formation Flight Test Objective

19

Flight Testing Plans

1) Validation of hardware system and formation flight control laws.

2) Wake visualization with a single aircraft.

3) Flight data collection for wind/gust/wake sensing with “Phastball” formation

flight.

Milestones

1) Outer-loop controller validation flight (Virtual Leader).

• Achieved on 07/22/2013;

2) Close formation flight of two “Phastball” aircraft (Leader-Follower).


3) Gust/wake measurements in close formation flight.


Nov. 16, 2012


November 14, 2013


updated for Phastball

1

u

(130)

Throttle

Switch1

Switch

surf

ilsw

u

Subsystem

U Y

UY

UY

UY

UY

Scope2

g3dl_s_3NT

OuterLoop

obes_v1

OBES1

pqrtp

ph_cmd

th_cmd

de

da

dr

Inner loop controller

Formation

Signals

Demux

Demux

0

[0 3 130]

Biases1

[0 1.1 0]

Biases

2

r

1

y

DeltaR

DeltaA

DeltaE

ThrottleThrottle

Outer Loop Controller • generates to track the virtual leader

• inputs: the states of the plane and the position of the VL

• based on NLDI control techniques

,ref ref

Inner Loop Controller • generates to track

• inputs:

• baseline design with LQ control techniques

,ref ref , ,el al R

p q r

Surface Deflections • 1 command for elevator

• 1 command for ailerons

• 1 command for the rudder

• 1 command for the throttle

Inputs • States of the AC

• States of the

Virtual Leader

Formation Flight: Control Laws Implementation

Nov. 16, 2012


November 14, 2013


Formation Flight Control Scheme

21

Avionics Software

• Matlab/Simulink tool chain for flight avionics software downloading.

• Matlab/Simulink RTAI C source files Linux compilation Executable

Nov. 16, 2012


November 14, 2013


2013 Flight Season Summary

22

Flight Testing Session # Date Aircraft

Data Set # Mission Comments

1 5/25/2013 Blue 1 Trim/ Innerloop tracking 0˚ roll and 2˚pitch 2 Smoke Plane was manually flown through a smoke screen for visualization of wing vortices 3 Outer Loop Green 1 Trim/ Innerloop repeat of blue1 flight, rudder oscillations, taped on camera, one choke on reciever wire

2 6/14/2013 Green 1 Inner Loop Favorable behavior seen during control activation Blue 1 OuterLoop Catastrophic Failure resulting in crash.

3 7/12/2013 Green 1 Innerloop / background Online VL Favorable behavior seen during control activation 2 Online VL outerloop FF control Forward distance appears to be negative when it should be positive. Red 1 Trim/ return to action Red put back into action after losing Blue.

4 7/22/2013 Green 1 inner loop Control on during straight legs 2 outer loop No unexpected errors. Clearance [10 0 0] 3 outer loop Vrt. Dist. Gain increased to 1.3 4 outer loop Successful virtual leader flight. Red 1 shake down 2 shake down

5 7/25/2013 Green/Red 1 inner loop Favorable behavior seen during control activation 2 red downlink green was recieving data while sitting on the ground and red flew 3 Formation Flight! red leader; green follower; Holding pattern ; Constant clearance 4 Formation Flight! Constant clearance 5 Formation Flight! Constant clearance

6 9/8/2013 Green/Red 1 Formation Flight! Varying clearance 2 Formation Flight! Constant clearance 3 Formation Flight! Varying clearance

7 10/12/2013 Green/Red 1 Formation Flight! Constant clearance 2 Formation Flight! Constant clearance 3 Formation Flight! Varying clearance 4 Formation Flight! Constant clearance

Nov. 16, 2012


November 14, 2013


2-Aircraft Formation Flight Experiments Summary

23

Flight # Mission Description

1 Formation Flight, Holding Pattern Forward Clearance: 50m



4 Close Formation Flight with Pilot

Adjustments

Varying Clearance:

Forward 24±12m, Lateral ±12m, Vertical ±12m

5 Close Formation Flight, Holding Pattern Forward Clearance: 12m


Adjustments

Varying Clearance:

Forward 24±12m, Lateral ±12m, Vertical ±12m


Vertical Clearance: -2m (corrective bias)




Adjustments

Varying Clearance:

Forward 24±12m, Lateral ±12m, Vertical -2m



Nov. 16, 2012


November 14, 2013


24

Flight Experiments

Video

Nov. 16, 2012


November 14, 2013


Steady State Performance Analysis: Straight Legs

25

FF Straight legs Clearance Max Err. Distance Mean Abs. Err.

Distance Mean Err. Distance Std. Dev.

avg. % wing span

Flight 1

Forward (m) 50 -6.112 2.623 -2.356 1.896 98% Lateral (m) 0 -5.615 2.011 -1.628 1.985 68% Vertical (m) 0 4.778 2.617 2.617 0.993 109% Magnitude (m) 50 9.577 4.216 3.879 2.919 162%

Flight 2


Flight 3


Flight 5

Forward (m) 12 2.068 0.533 0.494 0.486 21% Lateral (m) 0 -1.890 1.193 -1.050 0.695 44% Vertical (m) 0 3.088 2.391 2.391 0.386 100% Magnitude (m) 12 4.170 2.724 2.657 0.931 111%

Flight 7

Forward (m) 12 1.899 0.649 -0.499 0.596 21% Lateral (m) 1.2 0.551 0.184 -0.021 0.238 1% Vertical (m) 2 2.229 1.640 1.640 0.212 68% Magnitude (m) 12.2 2.979 1.773 1.715 0.676 71%

Flight 8


Flight 10


*Flights 4,6, and 9 are not analyzed because the formation geometry was varying during the flights.

** GPS error was not considered in the performance analysis.

Nov. 16, 2012


November 14, 2013


26

FF Turns Clearance Max Err. Distance Mean Abs. Err.

Distance Mean Err. Distance Std. Dev.

avg. % wing span

Flight 1

Forward (m) 12 -12.475 5.650 -7.177 5.220 299%

Lateral (m) 0 -22.371 8.048 -12.949 5.437 540%

Vertical (m) 0 9.051 4.209 5.481 2.247 228%

Magnitude (m) 12 27.166 10.696 15.786 7.865 658%

Flight 2


Flight 3


Flight 5

Forward (m) 12 1.986 0.762 0.729 0.445 30% Lateral (m) 0 3.438 2.394 2.394 0.524 100% Vertical (m) 0 9.485 3.960 3.960 1.052 165% Magnitude (m) 12 10.282 4.690 4.684 1.256 195%

Flight 7

Forward (m) 12 2.951 1.863 1.863 0.445 78%

Lateral (m) 1.2 4.177 3.180 3.180 0.469 132%

Vertical (m) 2 6.812 4.265 4.265 1.380 178%

Magnitude (m) 12.2 8.518 5.637 5.637 1.524 235%

Flight 8

Forward (m) 12 6.059 3.431 3.431 1.307 143% Lateral (m) 1.2 4.402 3.836 3.836 0.221 160% Vertical (m) 2 8.423 5.994 5.994 1.015 250% Magnitude (m) 12.2 11.271 7.900 7.900 1.669 329%

Flight 10

Forward (m) 12 3.338 0.949 0.818 0.885 34% Lateral (m) 1.2 4.512 3.561 3.561 0.479 148% Vertical (m) 2 11.391 8.718 8.718 1.585 363% Magnitude (m) 12.2 12.699 9.465 9.452 1.877 394%

Steady State Performance Analysis: Turns

Nov. 16, 2012


November 14, 2013


Sample Flight Data

27

Straights

Turns

Nov. 16, 2012


November 14, 2013


Outline






28

Nov. 16, 2012


November 14, 2013


29

Basic Strategy

• Leader: measurement of local wind field,

• Follower: measurement of local wind field + wake generated by the leader.

Outline

• 3D wind estimation using Unscented Kalman Filter (UKF).

• Cooperative estimation of wind field using UKF.

• Wake sensing in UAV formation flight (flight test results).

Cooperative Wind/Wake Estimation in Formation Flight

Nov. 16, 2012


November 14, 2013


30

Kinematic Equations for UKF Approach

• UKF was selected over EKF because of its effective linearization

technique and ease of implementation.

• Update equation: 𝑥 = 𝑈, 𝑉,𝑊,Φ, θ, ψ, 𝑤𝑥 , 𝑤𝑦, 𝑤𝑧 :

𝑈

𝑉

𝑊 =

𝑟𝑉 − 𝑞𝑊 + 𝑎𝑥𝑝𝑊 − 𝑟𝑈 + 𝑎𝑦𝑞𝑈 − 𝑝𝑉 + 𝑎𝑧

+ 𝐷𝐶𝑀(Φ, θ, ψ)𝑇00𝑔

• Measurement equation:

𝑉𝑝𝑖𝑡𝑜𝑡𝛼𝛽

=

𝑈

𝑡𝑎𝑛−1(𝑊

𝑈)

𝑠𝑖𝑛−1(𝑉

𝑈2+𝑉2+𝑊2)

,

𝑉𝑥𝑉𝑦𝑉𝑧

= 𝐷𝐶𝑀(Φ, θ, ψ)𝑈𝑉𝑊

+

𝑤𝑥𝑤𝑦𝑤𝑧

Wind Estimation Using Unscented Kalman Filter

Nov. 16, 2012


November 14, 2013



• Simulator sensor noise from “Phastball” UAV GPS/INS/ADS

• Compared results with a direct calculation method

» UKF mean: -0.0676 deg., std: 0.0836

» Direct mean: -0.2058 deg. , std: 0.4233

0 10 20 30 40 50 60 70 80 90-2

-1

0

1

2

3

4

5

6

7UKF Calculated Wind

Time (sec)

Tota

l W

ind V

elo

city (

m/s

)

Simulated

Calculated

Difference

0 10 20 30 40 50 60 70 80 90-2

-1

0

1

2

3

4

5

6

7

Time (sec)

Tota

l W

ind V

elo

city (

m/s

)

Direct Wind Calculation

Simulated

Calculated

Difference

Nov. 16, 2012


November 14, 2013



• Validation using flight test data and ground weather station

400 450 500 550 600 650-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3

3.5

Time (sec)

Win

d V

elo

city (

m/s

)

GPS/INS/ADS Wind States

wx

wy

wz

400 450 500 550 600 6500

1

2

3

4

Time (sec)

Tota

l W

ind S

peed (

m/s

)

Wind Estimation with Weather Station Reference

GPS/INS/ADS

WS Power Law

WS Uncorrected

400 450 500 550 600 650

-50

0

50

Time (sec)

Win

d D

irection (

deg)

GPS/INS/ADS

WS

UKF estimation of 3D wind. Horizontal planar wind speed and direction.

Nov. 16, 2012


November 14, 2013


Cooperative Gust Sensing with UAV Formations:

Simulation Results

Leader Wind

Estimation (UKF1)

Follower Wind

Estimation (UKF1)

Cooperative Wind

Estimation (UKF2)

Mea

n o

f

erro

r x (m./s.) 0.5138 0.4568 0.5500

y (m./s.) 1.1869 1.4173 1.1595

z (m./s.) 0.2518 0.2564 0.2280

Norm (m./s.) 1.3176 1.5111 1.3034

Std

. of

erro

r x (m./s.) 0.3254 0.3256 0.3428

y (m./s.) 0.8071 0.8761 0.8216

z (m./s.) 0.2263 0.2557 0.2270

Norm (m./s.) 0.8991 0.9690 0.9187

• UKF1 is to estimate the 3D wind field using only the local measurements.

Leader/follower wind estimation can be obtained separately using UKF1.

• UKF2 is a cooperative strategy to estimate the local wind field using both leader

and follower information. Wake predictions from the leader are added to the

measurement equations of UKF2.

• “1 minus cosine” gust profile and Phastball wake model is used in simulation.

Nov. 16, 2012


November 14, 2013


• Motivations

– How to utilize the wind information measured by the leader?

– How to compensate for the vortex turbulence during tight formation flight?

• On-Going Efforts

– Longitudinal dynamics is focused for the proof of the concept.

– A feed forward controller is being added to the current inner loop flight

controller using the developed gust/wake estimation.

– Initial work will be on the validation of the concept without the wake

effect.

– Given the estimated wake location and strength, how to design a new

controller for gust suppression control.

Gust Suppression with UAV Formations

Nov. 16, 2012


November 14, 2013


Wake Sensing: Flight Results

• Nose board provides:

– Static/dynamic pressure data from Pitot tube,

– Flow angles from two alpha vanes and one beta vane.

• Weather station collects wind speed and direction data on the ground.

• The air data system was calibrated on a calm day.

Ground Weather Station Pressure sensor board Flow angle potentiometers

Nov. 16, 2012


November 14, 2013


Wake Sensing: Flight Results (Cont.)

• Wake experienced by the Follower (the left and right α vanes are 25

cm. apart laterally)

42.5 43 43.5 44 44.5 45-5

0

5

(d

eg)

Follower Aircraft Data

Right

Left

Difference

42.5 43 43.5 44 44.5 450

5

10

Pitc

h (d

eg)

42.5 43 43.5 44 44.5 45-5

0

5

A z (m/s2 )

Time (sec)

Nov. 16, 2012


November 14, 2013


Outline






37

Nov. 16, 2012


November 14, 2013


Conclusions

• Successfully achieved close formation flight (up to 5 b or ~12 m.) with 2

low-cost UAV research platforms. The developed formation flight

controller behaved desirably within ~1 m. standard deviation during

straight legs.

• Demonstrated that small sub-scale research aircraft (~25 lbs.) can

generate vortices strong enough to be sensed by the following aircraft

without being buried in the ambient wind turbulences.

• Showed initial advantages for cooperative wind/gust estimation and

suppression control with formation flight in simulation.

38

Nov. 16, 2012


November 14, 2013


Plans for Phase II

39

• “Phastball” platforms to be upgraded with higher quality sensors

including RTK GPS (1 cm.) and spatially distributed 5-hole Pitot tubes;

• Investigate the interactions between the ambient wind and wake-induced

vortices;

• Real-time estimation of the wake vortex center;

• Real-time cooperative gust sensing and control;

• Quantify the aerodynamic benefits of a dynamic ‘sweet spot’ following

close formation flight;

• Scalability analysis for different weight/classes of aircraft.

Nov. 16, 2012


November 14, 2013


Thank You!

http://fcrl.mae.wvu.edu

http://www2.statler.wvu.edu/~irl

http://fcrl.mae.wvu.edu/

http://fcrl.mae.wvu.edu/



NASA LEARN Program · • Conference paper presented at the 2013 AIAA GNC Conf. • Conference paper submitted to the 2014 ACC (details in next page). ... (Virtual Leader). • Achieved

Documents